Understanding Interactions between Biomolecules and Two-Dimensional Nanomaterials Using in silico Microscopes

Abstract

Two-dimensional (2D) nanomaterials such as graphene are increasingly used in research and industry for various biomedical applications. Extensive experimental and theoretical studies have revealed that 2D nanomaterials are promising drug delivery vehicles, yet certain materials exhibit toxicity under biological conditions. So far, it is known that 2D nanomaterials possess strong adsorption propensities for biomolecules. To mitigate potential toxicity and retain favorable physical and chemical properties of 2D nanomaterials, it is necessary to explore the underlying mechanisms of interactions between biomolecules and nanomaterials for the subsequent design of biocompatible 2D nanomaterials for nanomedicine. The purpose of this review is to integrate experimental findings with theoretical observations and facilitate the study of 2D nanomaterial interaction with biomolecules at the molecular level. We discuss the current understanding and progress of 2D nanomaterial interaction with proteins, lipid membranes, and DNA based on molecular dynamics (MD) simulation. In this review, we focus on the 2D graphene nanosheet and briefly discuss other 2D nanomaterials. With the evergrowing computing power, we can image nanoscale processes using MD simulation that are otherwise not observable in experiment. We expect that molecular characterization of the complex behavior between 2D nanomaterials and biomolecules will help fulfill the goal of designing effective 2D nanomaterials as drug delivery platforms.

Graphical Abstract

Introduction

Two-dimensional (2D) nanomaterials have expanded in usage and popularity since the discovery of graphene. Carbon-based nanomaterials, such as graphene and carbon nanotubes (CNTs), have been extensively studied for nanotechnology applications because of their mechanical stability,¹ attractive surface chemistries, ease of functionalization,² favorable electronic conductivity,³ and outstanding optical properties.⁴ Recent advances in graphene fabrication have broadened the application of graphene into areas such as low-dimensional transparent electrodes,⁵ optical sensors for nanoelectromechanical systems (NEMS),⁶ surface coating materials,⁷ biological probes,^8,9 and field-effect transistors.¹⁰ Due to graphene’s favorable properties and ease of synthesis, it is used in many industrial and consumer products and its applications continue to expand.

Beyond graphene, many non-carbon based 2D nanomaterials such as hexagonal boron nitride (h-BN) and molybdenum disulfide (MoS₂) nanosheets have met renewed research interest in the last decade. MoS₂ is a typical metal dichalcogenide containing weakly interacting layers in its crystal structure. Due to its outstanding tribological properties, MoS₂ has been used as a dry lubricant, such as in engines, for decades. Advances of recent nanotechnologies enable the fabrication of nanometer-sized MoS₂ with reduced dimensions, such as single-layer 2D MoS₂ nanosheets,¹¹ nanotubes (1D),¹² as well as fullerene-like nanoparticles (0D). The quantum confinement of MoS₂ nanomaterials yields appealing optical¹³ and electrical¹⁴ properties that have lead to many novel applications, including gas sensors,^15,16 ultra-sensitive photo-detectors,¹⁷ electrochemical hydrogen storage,¹⁸ and 2D field-effect transistors (FET).¹⁹ In addition, MoS₂ nanosheets might be an alternative 2D semiconductor material to graphene due to its finite band-gap electronic structure.

Remarkably, recent advances in nanofabrication have allowed researchers to integrate distinct 2D materials into van der Waals (vdW) heterostructures.²⁰ Many different 2D materials can be combined into one vertical stack, held together by van der Waals forces, yielding functional materials with diverse behavior and applications such as the unusual electronic properties of germanene on MoS₂²¹ as well as field-effect transistors (FET)²² and photo-responsive memory devices²³ formed by graphene-MoS₂ heterostructures. Furthermore, different 2D nanomaterials can be integrated together to form an in-plane heterostructure with controlled domain sizes, including seamlessly connected graphene and h-BN,²⁴ WS₂ and MoS₂,²⁵ MoS₂ and MoSe₂²⁶ as well as MoSe₂ and WSe₂.²⁷

Compared to the extensive studies of 2D nanomaterial electronic structures, 2D nanomaterial interactions with biological molecules have lagged behind but are receiving increasing attention in recent years. The urgency of research in this field arises from two factors. One factor is the growing concerns about biosafety and potential toxicity given the wide applications of these 2D nanomaterials. Experiments have revealed that graphene and CNTs can be internalized into the cytoplasm and nucleus.^28,29 Thus, interactions between carbon-based nanomaterials and biomolecules inside the cell can interfere with biological functions, resulting in cytotoxicity.³⁰ The second factor is that 2D nanomaterials, with appropriate size tuning and surface functionalization, are excellent candidates in many bionanotechnology applications such as biosensors³¹ and drug delivery vehicles.³² Ideally, as a nanomedicine, a functionalized 2D nanomaterial should be safe to host cells yet target disease-causing cells, bacteria, or virus as desired.

To summarize existing knowledge and advance the field, here we review previous theoretical studies of interactions between biomolecules and 2D nanomaterials, focusing on the molecular level comprehension. 2D nanomaterial biomedical applications, such as biosensors and drug carriers, remain limited due to our lack of knowledge of the complex behavior between 2D nanomaterials and biomolecules. Because the lengthscale of these biomolecules is smaller than the wavelength of light, elucidating molecular interactions is either costly or out of reach of current experimental techniques. However, complementary to experiments, computational modeling techniques including molecular dynamics (MD) simulations have been widely applied to investigate nanoscale processes.^33,34

With sophisticated force fields for describing molecular interactions, MD simulations have become a valuable resource for investigating the dynamic movements of biomolecules and nanomaterials using high performance computing (HPC) platforms. Force fields for classical molecular dynamics are typically developed for biomolecules in solution. To accurately model the bio-nano interface, new nanomaterial force fields compatible with existing biomolecular force fields are developed using both advanced quantum chemical and empirical methods. Different force fields for graphene and its derivatives, including modified van der Waals parameters and partial charge assignments for classical and polarizable force fields are discussed in references.^35,36 Additionally, for other 2D nanomaterials, such as gold and silver monolayers, force fields are developed to accommodate multiple distinct metal atom arrangements, such as the Au(111) and Au(100) surfaces, as well as the resulting bio-metal interfaces.^37,38

We arrange our review into four sections. First, we discuss the complex interactions between 2D nanomaterials and proteins. The elemental compositions and arrangement of a 2D nanomaterials play an important role in protein adsorption.³⁹ Similarly, protein secondary and tertiary structure as well as complex assembly may be modified upon binding to the material surface due to the strong adsorption interactions.⁴⁰ Proteins carry out various biological functions in the cell through protein-protein interactions (PPIs). An abnormal PPI can cause functional failure and lead to many diseases, including cancer⁴¹ and Alzheimer’s disease.⁴² Therefore, understanding 2D nanomaterial-protein interactions will aid the design of biocompatible nanomaterials that retain protein conformations and PPIs.

Second, we review studies focusing on 2D nanomaterials’ interaction with biological lipid membranes. Experiments have shown that graphene can penetrate the cell membrane.^29,43 We discuss two mechanisms of this membrane penetration behavior: extracting a large amount of phospholipids by graphene²⁹ and/or piercing a cell membrane by graphene’s sharp corners and jagged protrusions along the irregular edges.⁴⁴ We show that the graphene’s strong dispersive adhesion with lipids plays a dominant role for membrane piercing and lipid extraction. Surface functionalizations, including PEGylation, modulate, but do not altogether prevent this membrane penetration behavior.^45,46 We further discuss ion concentration effects on membrane penetration as well as favorable orientations of nanomaterial-membrane interactions.

Then, we review a few studies aiming to understand DNA’s interactions with 2D nanomaterials. Experiments have found that nanomaterials, such as MoS₂ have different affinities for single- and double-stranded DNA, yielding a sensor for sequence-dependent hybridization,⁴⁷ and a nanopore on the ultra-thin MoS₂ membrane can differentiate single nucleotides⁴⁸ with important implications for DNA sequencing. In addition to MoS₂, here we focus on the molecular level understanding of DNA’s interaction with various 2D nanomaterials as well as newly discovered 2D van der Waals heterostructures. The latter contains two distinct van der Waals materials (e.g. graphene and h-BN) resulting in different chemical potentials for an adsorbed ssDNA molecule and thus provides a versatile platform for manipulating the conformation of ssDNA.

Lastly, with modest understanding of the interactions between biomolecules and 2D nanomaterials, we further discuss the feasibility of various 2D nanomaterials as drug delivery platforms. 2D nanomaterials can effectively load drug molecules onto their large surface area through physical adsorption and specialized surface functionalization enables selective delivery of drug molecules to targeted cancer cells or bacteria.⁴⁹ We review studies on photothermal applications of 2D nanomaterials, cell internalization mechanisms, as well as circulation and clearance of these nanocarriers.

In sum, understanding interactions between biomolecules and 2D nanomaterials at the molecular level is crucial for mitigating potential toxicity and designing biocompatible 2D nanomedicines. Moreover, gaining insight into the mechanisms responsible for these interactions can greatly facilitate the discovery and development of novel drug delivery platforms for small molecules and biologics.

2D nanomaterial-protein interactions

Proteins are molecular machines that perform specific functions for life. There are antibodies that are produced by the immune system to protect our body from infection, enzymes that catalyze chemical reactions in cells, protein hormones that regulate biological processes, and so forth. The building blocks of proteins are twenty amino acids, each of which has a unique side chain that differs in size, charge, and geometry. They are joined together by peptide bonds, and the resulting polypeptide chain folds into a specific 3D structure depending on the interactions between the constituent amino acids as well as the interactions between these amino acids and other molecules in the environment. Most functional proteins are globular and soluble in water. The amino acids arrange the backbones and side chains driven largely by hydrophobic interactions, stabilizing the thermodynamically favorable and compact 3D structures.⁵⁰

Immobilization of proteins restricts their movement, and therefore improves protein stability and activity.^51,52 Benefiting from their unique physical and chemical properties, 2D materials are attractive immobilization substrates for proteins. The large surface area to volume ratio provides high protein adsorption capacity, and the atomic thickness offers mechanical flexibility for maximal contacts between the material surface and adsorbed protein. Other surface properties, such as functionality, crystallinity, and curvature also have strong impact on material-protein interactions and may be modified to modulate protein adsorption. In the following sections, we discuss a general mechanism of protein adsorption from the bulk solution onto 2D nanomaterial surfaces as well as nanomaterial and protein properties that influence protein adsorption.

Molecular mechanism of protein adsorption on 2D nanomaterial surfaces

A general molecular mechanism of protein adsorption onto 2D nanomaterial surfaces consists of three phases as identified by Penna et al. using large scale MD simulations of peptides and 2D nanomaterial surfaces in explicit solvent^53,54 (Figure 1).

Figure 1:

A general three-phase mechanism of peptide adsorption onto a 2D nanomaterial surface identified by Penna, Mijajlovic, and Biggs: (1) the biased diffusion phase, where the peptide in the bulk solution moves toward the surface driven by electrostatic interactions between interfacial water molecules or hydrophobic effect of nonpolar side chains of the peptide; (2) the anchoring phase, where the peptide is anchored to the second water layer through hydrogen bonding; (3) the lockdown phase, where the anchored peptide moves further into the first water layer and gradually rearranges itself to maximize the direct interaction with the surface,⁵³ with permission from the American Chemical Society.

The first stage is the biased diffusion phase where the protein moves from the bulk solution toward the material surface. Depending on the charge density of the material as well as the hydrophobicity and surface charge of the protein, diffusion can be driven by long-range electrostatic interactions between the water molecules adjacent to the surface and the water molecules around the protein⁵³ or by hydrophobic interactions between the interfacial water molecules and the nonpolar amino acids of the protein.⁵⁵ The former was observed on 2D metal surfaces, such as platinum and gold monolayers, where the interfacial water molecules are structurally oriented, creating a net negatively charged layer near the top surface that contributes to the electrostatic interactions. The latter was found on the surface of 2D layered graphene, graphite, where the hydrophobic interactions are more dominant than the electrostatic attraction.

The second stage is the anchoring phase, where the anchoring residues of the protein interact with the second water layer reversibly. For 2D metal surfaces, the anchoring residues are usually hydrophilic such as Ser⁵⁶ with anchoring driven by hydrogen bonding between their polar side chains and the water layer. Hydrophobic residues such as Phe and Met also have high anchoring propensity due to dispersion interactions arising from the electron-rich benzyl ring and sulfur atom, respectively. For graphene surfaces, hydrophobic and aromatic residues have high anchoring propensity, and anchoring is mainly driven by hydrophobic sequestration of the hydrophobic side chains from the water layer.

The third and last stage is the lockdown phase where the anchoring residues further interact with the first water layer and the anchored protein rearranges in a stepwise manner to gradually increase the number of direct contacts with the surface. Compared to the anchoring phase, the lockdown phase is slower as it requires formation of a cavity in the first water layer at the location adjacent to the anchoring residues, and the first water layer is more tightly bound to the surface than the second water layer. Depending on the adsorption strength between the material surface and adsorbed protein, the adsorbed protein may reversibly disengage the first water layer and attach to the surface during the lockdown phase.

2D nanomaterial properties

Based on different elemental compositions and arrangements, 2D materials possess a wide range of structural, physical, mechanical, electrical, chemical, thermal, and optical properties for diverse applications. Among these properties, stable aqueous dispersibility (physical), high biocompatibility and low toxicity (chemical) are critical for biomedical applications. In the following section, we discuss some of the important factors that influence these physical and chemical properties of 2D materials in the context of 2D nanomaterial-protein interactions.

Surface functionality effects

Colloidal stability of 2D nanomaterials under biological conditions largely depends on their aqueous dispersibility. Materials with poor aqueous dispersibility tend to form 3D aggregates that are undesirable for applications such as protein adsorption and drug delivery. Although altering environmental factors, such as pH or ionic strength of the solution, may improve dispersibility,⁵⁷ these changes are difficult to implement in physiological conditions. An alternative strategy is to modify the nanomaterial surface by covalent or non-covalent functionalization which modulates material properties including dispersibility.^58,59 Increasing surface oxidation, for example, leads to graphene oxide (GO) nanosheets being more dispersible than pristine graphene nanosheets.⁵⁸ However, the oxygen moieties at the surface of the GO nanosheet can transfer electrons with components of the mitochondrial electron transport chain, which produces reactive oxygen species and induces cytotoxicity in lung macrophages.⁶⁰ Moreover, introducing polar or charged functional groups on material surfaces may reduce the biocompatibility of the materials as these covalent modifications change the hydrophobicity and the direction of electric dipoles of material surfaces, which in turn affect protein conformation and orientation (Figure 2A).⁶¹

Figure 2:

Effect of 2D nanomaterial properties on protein adsorption. (A) Protein hydrophobin adsorbed on four different self-assembled monolayers (SAMs), terminated with the methyl (−CH3), hydroxyl (−OH), carboxyl (−COOH), and amino (−NH2) groups, respectively. These functional groups induce hydrophobic (h) and electric (e) dipoles of the protein that affect the conformation and orientation of the protein on the surface. Adapted with permission from ref.⁶¹ Copyright 2014 American Chemical Society. (B) Protein dissociation on a heterogenous GO surface over 400 ns of MD simulation. The protein assembly Aβ16–21 peptides are shown in green. Adapted with permission from ref.⁶⁷ Copyright 2019 American Chemical Society. (C) The chicken villin headpiece subdomain unfolds on a defective graphene nanosheet. The anchoring residue (Arg-14) interacts with the carboxyl groups at the vacancy defect, while the aromatic residue (Phe-17) binds sp² hybridized regions. Reproduced from ref.⁶⁸ (D) The Aβ1–42 peptide denatures on a boron nitride surface as the curvature decreases from (3, 3) boron nitride nanotube (BNNT) to boron nitride nanosheet (BNNS). Adapted with permission from ref.⁶⁹ Copyright 2021 American Chemical Society.

Compared to covalent functionalization, non-covalent functionalization allows for the tuning of material properties without disrupting their elemental compositions or arrangements. Ionic surfactants, such as lithium salt of 6-aminohexanoic acid (Li-AHA), improve aqueous dispersibility of MoS₂ by electrostatic repulsion.⁶² Non-ionic surfactants, such as poly(ethylene oxide)-poly(propylene oxide)-poly(ethylene oxide) (PEO-PPO-PEO) triblock copolymer, known by the trade name Pluronic, enhances exfoliation of graphene, MoS₂, and h-BN by steric stabilization⁶³ and was found to increase cellular uptake of graphene and GO nanosheets in biological solutions.⁵⁸ In addition to Li-AHA and Pluronic, proteins are ideal surfactants and can modify material properties. Bovine serum albumin (BSA) adsorbs onto the surfaces of graphene and MoS₂ nanosheets through benzene rings and disulfide functional groups, serving as an exfoliating and stabilizing agent to keep the materials from aggregating in water.⁶⁴ Moreover, it is known that upon contact with biological fluids like blood serum, material surfaces are quickly coated with plasma proteins termed the protein corona.⁶⁵ The protein corona has a critical effect on the biocompatibility and cytotoxicity of nanomaterials, as exemplified by surface coating of GO nanosheets with fetal bovine serum (FBS) proteins or BSA decreases their cellular uptake and cytotoxicity.⁶⁶

Surface crystallinity effects

In the standard material synthesis process, perfect crystalline solids are rare. Instead, material surfaces are usually inhomogeneous and have defects. For instance, a heterogenous GO nanosheet has extended unoxidized regions and highly oxidized regions on its basal plane. The unoxidized regions are sp² hybridized, while the oxidized regions contain agglomeration of hydroxyl and epoxy groups. Such different surface polarity induces unfolding and dissociation of adsorbed proteins as their aromatic and nonpolar side chains prefer binding to the unoxidized regions through π-π stacking and hydrophobic interactions, and the polar side chains and backbones prefer interacting with the oxidized regions through hydrogen bonding (Figure 2B).⁶⁷ A similar effect was observed on a defective graphene nanosheet where carboxyl groups are attached to the carbon atoms at vacancy defects mimicking oxidation at the defects. These carboxyl groups interact with positively charged amino acids that serve as anchors in the anchoring phase of protein adsorption and facilitate protein denaturation (Figure 2C).⁶⁸

Surface curvature effects

Several studies have used MD simulations to investigate protein adsorption and conformational transitions on the surface of 2D nanomaterials of different curvatures. The modeled surfaces are usually 2D planar nanosheets and their corresponding nanotubes with various cross section diameters that mimic different curvatures. In general, the adsorption capacity increases as the surface curvature decreases and likewise the van der Waals interaction energy between the material and protein becomes more favorable with the decreasing curvature.^70,71 Moreover, different surface curvatures affect adsorption capacity, which in turn impacts the conformation of adsorbed proteins. A planar surface provides the highest adsorption capacity and tends to facilitate the structural transition of proteins to random coils and turns that maximize the contact area with the surface. On the other hand, a surface with a higher curvature has a lower adsorption capacity, and the adsorbed proteins are stable enough to retain their conformations (Figure 2D).⁶⁹

Other non-graphene 2D nanomaterials

Besides graphene, many other 2D nanomaterials have been investigated for their interactions with various proteins. For example, 2D molybdenum disulfide (MoS₂) nanosheet interacts only weakly with the first 17 residues of the huntingtin protein and thus the protein secondary structure is preserved.⁷² However, the same protein denatures on a graphene nanosheet due to the stronger protein-graphene interaction than the intra-protein interaction.⁷² Interestingly, on the surface of a black phosphorus monolayer, the villin headpiece sub-domain not only maintains its secondary structure but also diffuses quickly along the zigzag direction (i.e. an anisotropic effect).⁷³ Similar to graphene, graphyne is another carbon-based 2D nanomaterial that interferes with protein-protein interactions due to its hydrophobicity.⁷⁴

Protein adsorption on hexagonal boron nitride (h-BN) remains to be fully characterized. However, it is known that h-BN yields a stronger van der Waals attraction for an absorbed biomolecule than graphene.⁷⁵ Thus, a protein is likely to denature on a h-BN nanosheet.

Protein properties

Protein adsorption on material surfaces is a complex process. In addition to the surface properties of materials, the physicochemical properties of proteins and the external parameters of the environment, such as temperature, pH, and ionic strength of the solution, can influence protein adsorption. A comprehensive overview of different factors affecting protein adsorption has been discussed in a recent review.⁷⁶ In the following, we focus on the effects of protein sequence, conformation, orientation, and assembly on 2D nanomaterial-protein interactions.

Protein sequence, conformation, and orientation effects

Amino acid composition and sequence determine protein structure. Even one amino acid substitution may lead to large 3D conformational changes,^77,78 and different molecular conformations can affect their binding affinities to the material surfaces.^79,80 Therefore, by engineering protein sequence, one can modulate their conformation, orientation, as well as adsorption strength on material surfaces.⁸¹

The protein G_B is the immunoglobulin G (IgG) antibody-binding domain of protein G. Immobilized protein G is commonly applied as a regenerable immunosensor for IgG detection, and immobilized protein G along with bound IgG also serves as an immunosensor for antigen detection.^82,83 The 56-residue protein G_B folds into one α-helix and four β-sheets. Due to favorable π-π interactions between the aromatic side chains and the graphene surface, the protein has strong attraction with the surface that leads to denaturation (Figure 3A). However, with alanine substitutions for two hydrophilic residues (Gln32 and Asn35, see mutations in the protein sequence shown in Figures 3A and 3B) that have the smallest number of native contacts on the α-helix, the engineered protein G_B maintains its conformation on graphene (Figure 3B).⁸⁴

Figure 3:

Effect of protein properties on its adsorption onto a graphene nanosheet. (A) MD simulation results of the wild-type protein G_B adsorbed onto the graphene surface. The α-helix interacts with the surface and the protein loses its secondary and tertiary structures. (B) Adsorption of the protein G_B with two amino acid substitutions in the α-helix onto the graphene surface. One of the β-sheet interacts with the surface and the protein maintains its overall conformation. A and B are adapted with permission from ref.⁸⁴ Copyright 2019 American Chemical Society. (C) Time-dependent contact and distance analysis of a graphene insertion process into a hydrophilic protein-protein interface. (Top) Numbers of non-hydrogen atoms at the tubulin dimer–dimer interface (N_{α1β1−α2β2}) and at the dimer–graphene interface (N_α1β1−g and N_α2β2−g), and (bottom) time-dependent lateral distances between two dimers. (D) Snapshots of the graphene inserting into a hydrophilic protein-protein interface over simulation time points. C and D are reproduced from ref.⁸⁶ (E) Snapshots of the graphene inserting into a hydrophobic protein-protein interface over simulation time points, reproduced from ref.⁸⁷

In addition, protein secondary structures demonstrate varied degrees of conformational changes on graphene surface, with the WW domain (a β-sheet-rich protein) being more stable than the λ-repressor (a α-helix-rich protein). The former retains most of the β-sheet contents, while the latter loses helical contents especially at the protein-graphene interface, suggesting that the graphene nanosheet has a higher binding affinity for α-helices.⁸⁰ This observation indicates that the stability of a protein on graphene might depend on its orientation. Due to the strong van der Waals interaction between a protein and graphene, the protein may adopt multiple binding poses with different surface residues in contact with graphene, resulting in different orientations of the protein on graphene. For example, the different orientations of hydrophobin shown in Figure 2A could occur on the same graphene surface. Similarly, the protein G_B can maintain its overall conformation when one of the β-sheets interacts with the surface but loses its secondary and tertiary structures when the α-helix is at the interface (Figure 3A–B). The BBA protein (a protein that consists of one α-helix and two β-sheets) is not stable on the graphene surface, but the extent of denaturation largely depends on the protein orientations where the interfacial residues are different.⁸⁰ In general, aromatic residues have strong π-π interactions with the graphene surface, therefore, when there are aromatic residues at the interface, the protein has a higher tendency to be unfold. The binding strength of the aromatic residues is positively correlated with the polarizability of their aromatic motifs, where Trp binds the graphene nanosheet most strongly, followed by Tyr, Phe, and His.⁸⁵

Protein assembly effects

Depending on protein-protein interfaces, protein assemblies interact with graphene nanosheets differently. For a hydrophilic interface, such as the interface between tubulins of adjacent protofilaments that associate laterally to form microtubules, the assembly is stabilized by salt bridges between water-exposed side chains of opposite charge, Lys and Asp/Glu. Although it is more difficult for the graphene nanosheet to insert into a hydrophilic interface than to a hydrophobic one (see below), the strong dispersion force between the side chain of Lys and the graphene nanosheet breaks the salt bridges at the interface and initiates the graphene insertion as the graphene pulls the tubulins of a protofilament away from the tubulins of another protofilament (Figure 3C–D).⁸⁶

On the other hand, for a hydrophobic interface where the protein complex is stabilized by hydrophobic interactions, such as the dimer structure of the C-terminal DNA-binding domain of human immunovirus-1 integrase, the graphene nanosheet inserts directly into the interface⁸⁷ within merely a few nano-seconds (Figure 3E). Because the graphene surface is more hydrophobic than the interface (mainly formed by hydrophobic residues Leu242, Trp243, Ala248, Val250, Ile257, and Val259), this insertion is driven by favorable van der Waals interactions between the nonpolar residues at the interface and the graphene surface. The components of the disrupted protein complex then move independently on either side of the graphene nanosheet while maximizing their direct contacts with the graphene surface and minimizing the unfavorable interaction between the graphene surface and water (Figure 3E). A similar process was observed later between peptide aggregates and the graphene nanosheet.⁸⁸ The peptides gather to form an aggregate with a hydrophobic core, which is interfered by the graphene nanosheet and disassembles on the graphene surface.

2D nanomaterial-membrane interactions

Cellular membranes represent a challenge and an opportunity for biomedical applications of nanomaterials. Membranes are composed of amphiphilic phospholipid molecules which have both a polar head group (or a charged one for bacterial membranes), and a nonpolar, hydrophobic tail. These double-sided molecules align and orient themselves in polar solvents such as water so that the polar head groups are solvent exposed, while the hydrophobic tails are hidden from solvent either in a lipid bilayer such as a cell or liposome, or in a globular micelle. Membranes play important roles for the cell, including separation and containment of the living processes within the cell from the outside milieu. A large amount of biological 2D nanomaterial research has been devoted to exploiting this role of the membrane for biomedical purposes. Though many-sided, the primary reason for this is that graphene oxide (GO), the most studied 2D nanomaterial with membranes,⁸⁹ robustly penetrates the cell membrane and causes cell lysis.^29,44 This penetration and lysis behavior is most destructive in GO with oxygen contents less than 20%²⁹ and has been observed in both Gram-negative and Gram-positive bacteria,^90,91 as well as eukaryotes^92,93 including human cell lines.^94,95 This functionality is unique to 2D nanomaterials and stands in stark contrast to 0D and 1D nanomaterials which can be non-destructively internalized by cells.⁹⁶ The therapeutic implications of this penetration and lysis behavior are numerous, but mostly revolve around antibacterial and antimicrobial applications,^97,98 as well as selective medicinal applications such as antitumor drugs.⁹⁹ These applications currently remain in their infancy and require additional examination of GO nanosheet-membrane interactions and GO nanosheet-mediated toxicity. Here, we first review the current understanding of GO nanosheet-membrane interactions, focusing on mechanisms and modulation of these interactions through surface passivation and ion concentration. Next, we briefly review membrane penetration model theory and lastly discuss non-GO nanosheet-membrane interactions.

Interactions between graphene oxide and biological membranes

Membrane lysis mechanism

The mechanism of GO-membrane lysis itself has remained under debate since it was discovered.^29,44 Principally, two physical mechanisms were put forth, a lipid extraction mechanism and a piercing or blade-like mechanism. The lipid extraction mechanism observed by Tu et al.²⁹ suggested that the GO sheets inserted into the membrane and phospholipids diffused along the graphene sheets outside of the membrane. The lipids diffuse along the GO surface due to thermal fluctuations and strong dispersion forces between the GO sheets and lipid tail groups, see Figure 4A–B. Theoretically, this interesting phenomenon can be characterized as the complete wetting of graphene by membrane lipids.¹⁰⁰ The piercing mechanism, observed by both Li et al.⁴⁴ and Tu et al.²⁹ suggested the rigid GO sheets ‘sliced’ into the membrane and thereby caused cell lysis due to missing sections of membrane. Both mechanisms have been supported by experiments since they were put forth. Pham et al. found that GO-mediated cell death is proportional to edge density, which supports the piercing mechanism of cell lysis.¹⁰¹ In contrast, Zucker et al. found that membrane loss is proportional to GO surface area, rather than edge density, which supports the lipid extraction mechanism.¹⁰² However, we should note that so far, the experimental methods of these works and others have been heterogeneous and non-standardized, which makes cross-experiment comparisons treacherous. Standardized synthetic systems such as the lipid-filled nanopore system by Luan et. al,¹⁰³ Figure 4C, have been proposed to study GO nanosheet-lipid interactions, but have not been widely adopted. We also note that the lipid extraction theory requires a knife-like insertion at least partially into the cell membrane for lipids to diffuse across the GO sheet and be extracted from the membrane. Likewise, Duan et al.¹⁰⁴ found that when multiple GO sheets are inserted in close proximity in the cell, their lipid extraction mechanisms synergize to form perforations and pores in the cell membrane, Figure 4D. Hence, the two mechanistic theories of GO membrane lysis are not mutually exclusive and are likely dependent on GO fabrication, cell type, and environmental factors.

Figure 4:

2D graphene nanosheets penetrate cell membranes and compromise membrane integrity. (A) TEM image showing graphene nanosheets inserting into E. coli cells.²⁹ (B) E. coli lipid extraction mechanism observed via MD simulations.²⁹ (A-B) Reprinted by permission from Springer Nature: Nature Nanotechnology, ref,²⁹ 2013. (C) Simulations of a lipid-filled nanopore investigates lipid extraction in a controlled and quantifiable manner. Note that this model system uses POPC lipids found in eukaryotic membranes, but does not incorporate protein corona effects, see more in text. Reprinted with permission from ref.¹⁰³ Copyright 2017 American Chemical Society. (D) Membrane pore formation is observed when a high density of graphene sheets penetrate a eukaryotic cell membrane.¹⁰⁴ Similar to (C), protein corona effects are not included in ref.¹⁰⁴

2D surface coating effects

A popular way to reduce nanomaterial toxicity is via surface passivation: attaching a biologically inert molecule onto the nanomaterial surface to decrease harmful biological interactions. GO nanosheet surface passivation has been shown to attenuate, but not prevent membrane lysis as investigated by both experiment and molecular simulation. One biologically inert molecule that has been well studied for this purpose is polyethylene glycol (PEG), which has been successfully used to lower the toxicity of nanomaterials such as Quantum Dots.^105,106 Unfortunately, PEGylated GO nanosheets are still found to negatively interfere with cellular physiology. Luo et al.⁴⁵ found that PEGylated GO nanosheets caused an elevated macrophage cytokine response without being internalized by the cell. These PEGylated GO nanosheets were still able to penetrate the cell membrane and cause lipid extraction, despite the large hydrophilic PEG molecules adsorbed on the surface, Figure 5A. Interestingly, Luo et al. found that when GO nanosheets laid parallel onto the membrane, the PEG molecules could insert and anchor into the cell membrane Figure 5B, presenting another possible threat to membrane integrity. We note that PEGylated MoS₂ nanosheets do not disrupt the membrane as severely as PEGylated GO nanosheets,⁴⁶ implying that GO physicochemical properties are still intrinsically contributing to membrane disruption even when PEG passivated. From gene screening analysis and antibody blocking assays, Luo et al.⁴⁵ found that several membrane protein genes were dysregulated in macrophages after PEGylated GO exposure, but integrin β₈ was identified as the causative role for enhanced cytokine production. GO-mediated integrin activation was also discovered independently to occur in GO-exposed A549 lung cancer cells, specifically activating the integrin–FAK–Rho–ROCK pathway and leading to cell cytoskeleton disruption.¹⁰⁷ Integrin proteins are large transmembrane (TM) proteins that maintain a multitude of physiological roles including cell adhesion and signaling.^108,109 Follow-up GO-mediated integrin activation work by Chen et al¹¹⁰ found that lipid extraction by GO nanosheets caused the β₈ TM helix of integrin to be pulled away from the α_v TM helix, signifying a potential disruption mechanism of integrin activation by GO interactions with the TM helices, Figure 5C. We lastly note that even without surface passivation, GO nanosheets in physiological environments will adsorb a protein corona before cellular interaction, and these bound molecules will likely affect GO-membrane interactions. For example, Duan et al. ¹¹¹ found that when BSA proteins were first adsorbed onto GO nanosheets, GO nanosheet membrane penetration was weaker and cellular uptake of GO was reduced in comparison to bare GO. Additionally, Xu et al.¹¹² found that pristine and surface-passivated GO nanosheets still adsorbed IgG antibodies from serum, leading to differential membrane disruption effects in macrophages. The protein corona is not limited to BSA or IgG; instead the corona of proteins adsorbed onto GO sheets is time-dependent, with common, low-affinity proteins termed the soft corona adsorbing rapidly onto the GO surface, while more rare, higher-affinity proteins termed the hard corona eventually outcompete the soft corona and adsorb onto the GO surface.^113–115 Ultimately, any surface molecule on GO nanosheets, whether chemically bound or adsorbed, will affect GO-membrane interactions, and will require careful study to understand in vivo effects.

Figure 5:

Advanced study of 2D GO nanosheets reveals PEGylated GO still extracts membrane lipids, membrane-inserted graphene interferes with transmembrane proteins, and GO-membrane interactions are mediated by ion concentration. (A) Pristine GO and PEGylated GO simulations reveals that PEGylated GO still penetrates the cell membrane, although to a lesser extent than pristine GO. Simulation configurations of nanosheets at the same timestep show faster insertion of pristine GO (left) than PEGylated GO (right).⁴⁵ (B) PEGylated GO binding a membrane in the parallel configuration, with a PEG molecule inserting into the membrane.⁴⁵ (C) Lipid extraction by membrane-inserted graphene interferes with membrane proteins, shown here activating integrin α_vβ₈.¹¹⁰ (D) Increasing ion concentration causes enhanced GO nanosheet membrane penetration. The ions are thought to screen unfavorable electrostatic interactions between GO and membrane lipids, causing enhanced membrane penetration. Reprinted with permission from ref.¹¹⁸ Copyright 2015 American Chemical Society.

Ion concentration effects

GO nanosheet-membrane interactions are affected by ion concentration and valency. Though exact GO electrostatic properties are dependent upon fabrication, generally GO nanosheets are negatively charged^116,117 and must overcome electrostatic repulsion with polar or negatively-charged lipid head groups to interact with the membrane. Liu et al.¹¹⁸ found that GO nanosheets deposited faster onto negatively charged DOPC synthetic bilayers with increasing concentration of NaCl as well as CaCl₂, Figure 5D. The authors suspected that the increase in salt concentration screens the electrostatic clashes between the lipid head groups and the GO nanosheet, allowing for faster adsorption. Luan et al.¹¹⁹ found that when divalent Ca²⁺ ions bound to phospholipid headgroups, they caused a charge inversion at the membrane interface, resulting in a positively charged surface. This charge inversion provides theoretical support to the hypotheses of Liu et al., explaining how negatively-charged GO nanosheets increase membrane adsorption with increasing ion concentration.¹¹⁸ As seen with other nanomaterials,¹²⁰ however, any form of surface passivation may modulate this ion-dependent behavior.

Nanosheet-membrane orientation

In nanosheet-membrane interaction studies, there are generally only two orientations which are modeled: the perpendicular configuration, where the nanosheet penetrates the membrane at a 90° angle, Figure 5A, and the parallel configuration, where the nanosheet is face down on the lipid bilayer, Figure 5B. The perpendicular configuration is supported by theory as being the main penetration configuration due to splay and membrane tension energies.¹²¹ Splay is the capacity of the membrane lipid tails to reorient themselves and interact with lipid head groups as well as the opposite membrane leaflet.¹²² Nanosheet membrane penetration at an angle rather than perpendicular, would be “forced” back to the perpendicular configuration (the minimum free energy state) by splay and membrane tension energies. On the other hand, parallel configurations, where the nanosheet lies flat on the lipid head groups, is driven by membrane bending and tension energies.¹²¹ This theory largely explains why, in modeling studies of GO nanosheets and other 2D nanomaterials, typically only the perpendicular and parallel configurations are modeled: other configurations with the nanosheets inserted at an angle are not minimum free energy states.

Interactions between Non-graphene 2D nanomaterials and biological membranes

In addition to GO nanosheets, other 2D nanomaterial-membrane interactions have been investigated, including graphyne, MoS₂, WS₂, and h-BN among others. Graphyne, an sp¹ and sp²-bonded carbon nanomaterial, exhibits reduced membrane lipid extraction in comparison to GO,¹²³ thought to stem from the reduced carbon density of graphyne (Figure 6A). MoS₂ and WS₂, both long used as industrial lubricants,¹²⁶ can penetrate membranes and cause cytotoxicity.^{118,127–129} MD simulation reveals that MoS₂ extracts lipids from biological membranes, but at a rate that is an order of magnitude slower than graphene¹²⁴ (Figure 6B). However, though both MoS₂ and WS₂ disrupt the cell membrane, their cytotoxicity profiles are distinct. MoS₂ antibacterial properties are caused by a synergy of reactive oxygen species and cell membrane penetration.¹²⁸ In comparison, Liu et al.¹²⁷ determined that WS₂ antibacterial properties are not caused by reactive oxygen species and hence are primarily caused by membrane disruption. h-BN is another 2D nanomaterial with a long history of use as a lubricant as well as a makeup additive.¹²⁶ Despite its cosmetic application, it was found from both experiment and modeling,^125,130 that boron nitride nanosheets penetrate red blood cell membranes and extract lipids (Figure 6C), similar to the lysis mechanism of low-oxygen content GO.

Figure 6:

MD simulations of lipid extraction by 2D non-graphene nanomaterials. Snapshots illustrate the dynamic lipid extraction process for A) graphyne (Reprinted with permission from ref.¹²³), B) MoS₂ (Reprinted with permission from ref.¹²⁴), and C) h-BN (Reprinted with permission from ref.¹²⁵).

2D nanomaterial-DNA interactions

The ability to manipulate and analyze single-stranded DNA (ssDNA) is important to all bionanotechnologies such as nanopore sensors¹³¹ that promise to detect gene variations associated with various diseases as well as the transport and delivery of ssDNA. Generally, ssDNA in solution is in a coiled three dimensional (3D) conformation that requires unfolding to an extended state before performing sequence analysis. Due to the strong van der Waals adsorption of ssDNA on 2D nanomaterials,^132,133 the 3D ssDNA conformation is reduced to 2D on the nanomaterial surface. For a carbon nanotube (CNT) about 1 nm in diameter, the favorable van der Waals interaction can guide a coiled ssDNA molecule to enter the CNT and form a linear conformation.¹³⁴ Recently, nearly one dimensional conformations of ssDNA can be also achieved on specially engineered 2D nanomaterials.⁷⁵ Similarly, many 2D nanomaterial-based bionanotechnologies can potentially be extended to scrutinize single-stranded RNA (ssRNA), such as COVID-19 mRNA vaccines.¹³⁵

Most experimental work of DNA’s interaction with 2D nanomaterials is driven by the verification of recently developed theories for 2D polymers.¹³⁶ More recently, the potential application of 2D nanomaterials for transporting, sensing, and sequencing DNA have become highly attractive and have yielded a plethora of interesting results. For example, graphene nanosheet with a nanometer sized hole, i.e. nanopore, holds promise for the next generation low cost and high throughput DNA sequencing platform, because the nanopore’s atomic thickness is comparable to the spacing between two neighboring nucleotides in ssDNA, spatially permitting only one nucleotide inside the pore at any time for sensing. Despite in vitro studies, the implication of 2D nanomaterials’ interaction with ssDNA in a physiological environment warrants further studies for bio-safety. For that purpose, here we review current theoretical understandings of DNA’s interaction with 2D nanomaterials including graphene, hexagonal boron nitride (h-BN), and molybdenum disulfide (MoS₂).

ssDNA stretching on 2D nanomaterials

As demonstrated in experiment, 2D nanomaterials can be stacked on top of each other to form heterostructures²⁰ thanks to the strong van der Waals interaction on their surfaces. One natural example is graphite that is formed by the stacking of many graphene layers. More importantly, different types of 2D nanomaterials can also be stacked on top of each other, such as germanene on MoS₂²¹ and graphene on MoS₂.^22,23 In addition, two distinct 2D nanomaterials with similar lattice constants, such as graphene and h-BN, can be seamlessly connected laterally, i.e. side-by-side, forming in-plane heterostructures. Because of the strong π-π stacking with DNA bases, these so called van der Waals heterostructures have been proposed to manipulate, transport, and/or sense DNA.^75,137

It is well known from experiment that ssDNA prefers to adsorb and unfold or “lie down” onto the graphene surface.¹³⁸ Previous MD simulations⁷⁵ revealed that in 0.15 M KCl electrolyte, the intra-ssDNA π-π stacking between neighboring DNA bases was outweighed by the π-π stacking between DNA bases and graphene (Figure 7A). Similarly, ssDNA can rest on h-BN through the same type of interactions (Figure 7B). The π-π stacking of DNA bases with various 2D nanomaterials can be ranked as following: MoS₂ < graphene < h-BN.^75,137 For an in-plane van der Waals heterostructure formed side-by-side by graphene and h-BN nanosheets (Figure 7C), the defect-free junction permits the smooth diffusion of ssDNA across the interface. Due to the stronger interaction between ssDNA and h-BN, energetically, ssDNA prefers to stay on h-BN surface, which yields an one-direction drift from the graphene surface to the h-BN surface.⁷⁵ Additionally, one can potentially apply this in-plane heterostructure to concentrate ssDNA. For example, a large amount of ssDNA initially evenly distributed on a large graphene domain could eventually drift to a small h-BN domain, yielding a high concentration of ssDNA localized on the h-BN domain.

Figure 7:

ssDNA conformations on 2D nanomaterials. (A) ssDNA conformation on graphene. (B) ssDNA conformation on h-BN. (C) Adsorbed ssDNA near the graphene/h-BN junction. (D) Adsorbed and stretched ssDNA on a graphene/h-BN heterostructure. Carbon atoms in graphene are colored in gray; boron and nitrogen atoms in h-BN are colored in pink and blue respectively; ssDNA is in the van der Waals sphere representation with each nucleotide type colored differently. This figure was adapted from ref.⁷⁵

Notably, Luan et al.⁷⁵ found that a normally disordered and coiled ssDNA molecule can be spontaneously stretched into a linear conformation on an in-plane graphene/h-BN/graphene heterostructure as shown in Figure 7D, due to the above-mentioned differential interaction between ssDNA and nanomaterials. In this setup, a narrow stripe (∼ 2 nm) of h-BN is sandwiched between two graphene domains. From MD simulations, it was found that a circular shaped ssDNA on the graphene surface could quickly open up and form a linear conformation once it diffused to the h-BN stripe. Effectively, due to the atomic size difference, such a h-BN stripe in the in-plane heterostructure yields a surface channel for ssDNA with a width of about 2 nm and a depth of only 0.12 Å, reminiscent of much larger nanochannels used for confining and stretching dsDNA. However, the stretching of ssDNA on the h-BN stripe is due to the energy confinement, rather than a steric confinement.

The same above concept can be applied to other 2D in-plane heterostructures. For example, when connecting boronic graphene (BC3) or nitrogenized graphene (C3N) with a graphene nanosheet,¹³⁹ a short dsDNA fragment moves unidirectionally on the heterostructure due to the contrast of van der Waals interactions for the dsDNA on different substrates. Similarly, a disordered or denatured protein molecule can be stretched by a h-BN stripe sandwiched by two graphene domains.^140,141 In addition, a single-stranded RNA (ssRNA) in principle can also be manipulated on these engineered 2D in-plane heterostructures.

ssDNA transport through a nanopore in 2D nanomaterials

Mimicking a protein channel in the cell membrane, a nanometer-sized hole in a 2D nanomaterial can be used to transport biological molecules such as DNA and proteins and applied to next generation high throughput and low cost DNA and protein sequencing. Despite many years of investigations on the ssDNA’s translocation through graphene nanopores, an experimental demonstration has proven to be challenging. Recently, a theoretical work unveiled the myth of forbidden transport of ssDNA through a graphene nanopore.¹⁴² The challenge of ssDNA transport through a nanopore arises from both the notably inhomogeneous flexural rigidity of ssDNA bending around the atomically thin pore edge and the high energy cost due to the π-π stacking for a ssDNA base to desorb from the graphene surface before entering the pore.¹⁴² One simple solution is to drive the ssDNA molecule through a nanopore in a slightly thicker membrane, such as bilayer h-BN so that ssDNA bending is less profound.^142,143

In addition, van der Waals heterostructures can be used as a membrane with a finite thickness to contain a nanopore suitable for ssDNA translocation. As shown in the inset of Figure 8, Luan et al. ¹³⁷ used MD simulations to study ssDNA transport through a nanopore in a graphene-MoS₂ bilayer heterostructure. They found that a nanopore in this van der Waals heterostructure not only allowed ssDNA transport but yielded spontaneous ssDNA transport.¹³⁷ This striking phenomenon resulted from different van der Waals adsorption energies (or chemical potentials) between DNA-graphene and DNA-MoS₂ interactions.

Figure 8:

MD simulation of ssDNA transport through a graphene-MoS₂ heterostructure nanopore: the number (m) of transported nucleotides in the ssDNA molecule over time at various temperatures, 300 K (cyan), 350 K (orange) and 400 K (red and brown for two independent simulations). inset: the perspective top view of the simulation system. Atoms in the MoS₂ nanosheet are shown as van der Waals spheres (Mo:gray, and S:yellow); the graphene nanosheet (cyan) is in the bond representation; ssDNA is in the stick representation with each nucleotide type colored differently (A:blue; T:green; C:red; G:black); water is shown transparently; and K⁺ and Cl⁻ ions are colored in tan and cyan, respectively. This figure was adapted from ref.¹³⁷

It was discovered from MD simulations that DNA bases interact much stronger with graphene than with MoS₂.¹³⁷ Therefore, ssDNA can be driven from a MoS₂ nanosheet onto a graphene nanosheet. This behavior has been exploited for ssDNA transport using a 2D nanomaterial complex composed of a MoS₂ nanosheet on top of a graphene nanosheet with a nanopore extending through both layers. ssDNA initially adsorbed on the MoS₂ surface transits the pore and arrives at the graphene surface, driven energetically by the adsorption energy difference.¹³⁷ This spontaneous transport process is highlighted in Figure 8 showing the time-dependent number of transported nucleotides in ssDNA. After transport, all ssDNA bases were adsorbed on the graphene surface via π-π stacking. At T=350 K, interestingly, the transport number increased in a stepwise fashion and remained at each value for a finite amount of time, suggesting that at each step the ssDNA molecule was temporarily trapped due to the potential well resulting from the nucleotide-pore interaction. The large variation of trapping times at different steps suggests that the forward stepwise motion was thermally activated. When the temperature was switched to 400 K and 300 K, the transport process sped up and slowed down, respectively (Figure 8), confirming the thermally activated process.

Recently, using the same principle, the spontaneous transport of ssDNA through a nanopore in the vertically stacked BC3 and C3N heterostructure was also discovered.¹⁴⁴ Interestingly, the similarity of BC3 and C3N 2D-nanosheets yields close van der Waals interactions with ssDNAs, resulting in a weak driving force for ssDNA and consequently a slow transport of ssDNA.

Generally, ssDNA is driven through a solid-state nanopore using a biasing electric field, which yields a very fast ssDNA translocation limiting the sensing capability.¹⁴⁵ Here, the adsorption energy difference on both sides of a heterostructure provides a viable way to slowly transport ssDNA through a solid-state nanopore, facilitating molecular sensing and sequencing. On the other hand, for proteins, with their varied charge distributions along the polymer chain, the electric field driving method through a nanopore is not as effective. Thus, the van der Waals interaction difference discussed above becomes extremely important for driving a polypeptide chain through a solid-state nanopore.¹⁴⁶

ssDNA transport on a patterned 2D heterostructure

The nanopore structure described above allows the transport of a biological molecule in a 3D space from the cis. chamber on one side of the membrane to the trans. chamber on the other side. With the in-plane heterostructure, it becomes possible to design a molecule (or drug) dispenser, e.g. one-by-one transport of ssDNA through a 2D nanopore as shown in Figure 9.¹⁴⁷

Figure 9:

Transport of the ssDNA molecule through a 2-nm long 2D nanopore. (A-C) Snapshots of the simulation system at 281, 397, and 509 ns, respectively with a biasing voltage of 0.1 V. (D) The nanopore transport of ssDNA at various biasing voltages. Due to the periodic boundary condition, the ssDNA can transit the 2D nanopore multiple times along the x direction. This figure was adapted from ref.¹⁴⁷

MD simulations of ssDNA transiting a 2D graphene-(h-BN) nanopore reveal a “trapand-go” behavior as well as electric-field transport dependencies.¹⁴⁷ For a biasing voltage of 0.1 V, the ssDNA molecule was first trapped at the pore entrance (Figure 9A). After approximately 100 ns, the ssDNA molecule successfully entered and transited the short 2D nanopore (Figure 9B). Later, the ssDNA molecule diffused on the trans. h-BN domain (Figure 9C). This special trap-and-go transport process was observed several times during the simulation (the red line in Figure 9D). At V =0.2 V, a similar trap-and-go transport of the ssDNA molecule is shown in Figure 9D (the green line) with a faster transport speed. At an even higher biasing voltage (0.4 V, the blue line in Figure 9D), the ssDNA molecule transited the short nanopore with much less waiting (or trapping) time at the pore entrance because the entropy barrier is significantly tilted and reduced by the applied strong biasing electric field. Without a biasing electric field (V = 0 V), the ssDNA molecule was stopped by the entropy barrier and remained on the cis. h-BN domain (the black line in Figure 9D). Overall, for the 2D nanopore described above, each ssDNA molecule undergoes “trap” and “go” states in a biasing electric field. When the ssDNA is inside the 2D nanopore, its presence can be detected with electric signals.¹⁴⁷ Therefore, this designed 2D nanopore system provides a potential platform for precisely dispensing and sorting a certain amount of polymeric biologics to a target.

Generally, the strong van der Waals interaction between biomolecules and 2D nanomaterials enables the design of various nanostructures for manipulating, transporting, and sensing ssDNA, mRNA, and proteins. These heterogenous platforms have great potential for precisely handling biologics from source to targeted locations.

2D nanomaterials as drug delivery platforms

2D nanomaterials are versatile platforms for drug delivery.^32,148–151 Their large surface area to volume ratios and various surface functionalizations allow for effective loading of drug molecules through physical adsorption. In particular, 2D nanomaterials with hydrophobic surfaces can adsorb hydrophobic and aromatic therapeutics which are otherwise insoluble in water. Some 2D materials possess electromagnetic optical properties that result in photothermal heating triggered by near infrared light, enabling targeted therapeutic application. In this section, we discuss essences of 2D nanomaterial drug delivery, functionalization effects on drug absorption, targeted drug delivery mechanisms, photothermal activation, 2D nanomaterials with liposomes, cell internalization and circulation.

Essences of 2D nanomaterial drug delivery

GO nanosheet drug delivery stems from the investigation of carbon nanotubes for drug delivery^152–154 yet GO nanosheets have several inherent advantages over carbon nanotubes. In particular, the lower curvature and increased flexibility of GO nanosheets facilitates a broad, hydrophobic surface that readily binds aliphatic small molecule drugs and forms π-π stacking interactions with aromatic groups. The first examples of this behavior used PEGylated GO nanosheets (NGO-PEGs) to deliver antitumor drugs doxorubicin (DOX)¹⁵⁵ and topoisomerase inhibitor SN38¹⁵⁶ into cancer cell lines (Figure 10A). DOX is an effective anticancer drug that intercalates dsDNA and has been widely used in chemotherapy. SN38 is an intermediate metabolite of a clinical prodrug CPT-11 and is responsible for CPT-11 antitumor activity in colon cancer chemotherapy.¹⁵⁷ Unlike DOX, SN38 is insoluble in water but adsorbs onto the NGO-PEG surface by π-π stacking and hydrophobic interactions, leading to a NGO-PEG-SN38 complex that is water soluble and exhibits high potency for killing cells in a human colon cancer cell line HCT-116 in vitro.¹⁵⁶ In both studies, the therapeutics adsorbed noncovalently onto the NGO-PEGs and exhibited similar antitumor effects to administration of the drugs alone, while NGO-PEG-SN38 exhibited enhanced antitumor effects in comparison to NGO-PEG alone.¹⁵⁶

Figure 10:

2D nanomaterials for drug delivery. (A) (Left) Schematic illustrations of Doxorubicin (DOX)¹⁸² and SN38¹⁵⁶ loading onto PEGylated GO nanosheets (NGO-PEG). (Right, top) Relative cell viability as compared to untreated control of HCT-116 cells incubated with CPT-11, SN38, and NGO-PEG-SN38 at different concentrations. NGO-PEG-SN38 showed similar toxicity as SN38 in DMSO and much higher potency than CPT-11. (Right, bottom) Relative cell viability of HCT-116 cells after incubation with NGO-PEG with and without SN38. Plain NGO-PEG exhibited no obvious toxicity even at very high concentrations. Adapted with permission from ref.¹⁵⁶ Copyright 2008 American Chemical Society. (B) 2D covalent organic nanosheet (CON) conjugated with folic acid (FA) molecules for targeted delivery of the anticancer drug, 5-fluorouracil (5-FU). Reproduced with permission from ref.¹⁶⁴ Copyright 2017 American Chemical Society. (C) Photothermal drug delivery mechanism. Near infrared (NIR) light travels through biological tissues and heats a GO nanosheet, causing drug release. Reprinted from ref.,¹⁷² Copyright 2020, with permission from Elsevier. (D) 2D nanomaterial-mediated drug delivery in liposomes. GO nanosheets coat the liposomes and are irradiated with infrared light. The heated nanosheets cause drug release through heat-induced liposomal disruption.¹⁷⁶

Functionalization effects on drug adsorption

Functionalization of GO nanosheets results in complex drug adsorption. Mahdavi et al.^158,159 studied GO nanosheet PEGylation and oxidation effects on DOX absorption and observed that overall, decreasing PEG density and increasing surface oxygen density (up to ∼30%) increases DOX adsorption. PEGylation effects were also found to be PEG chain length dependent, with DOX molecules increasingly interacting and adsorbing onto longer PEG chains, rather than the GO surface.¹⁵⁹ It is unclear how this competitive adsorption behavior between the GO nanosheet and PEG molecules affect DOX drug delivery. Smaller functional groups also impact drug adsorption. Chen et al.¹⁶⁰ found that oxidation of the pristine graphene surface leads to more favorable electrostatic and hydrogen bonding patterns with the drug molecules primidone, pregabalin, and bortezomib. However, favorable interactions with the oxygen-containing groups also result in reduced drug unloading from the GO nanosheet, both in solvent as well as in the membrane leaflet,^160,161 indicating the need for a balance in the degree of oxidation of the GO nanosheet for adequate drug delivery.

Targeted drug delivery mechanisms

In addition to favorable drug adsorption, effective drug delivery requires 2D nanomaterials to release drugs at specific locations of therapeutic action. Sun et al. ¹⁵⁵ and others^149,162 have found that acidic environments, like the tumor microenvironment, accelerate drug release from GO nanosheets, indicating a potential chemical-mediated targeting mechanism. A popular biological targeting mechanism for 2D nanomaterial drug delivery is to attach cell specific binding moieties to the material surface so that the 2D materials will selectively bind certain cells. In the same study, Sun et al.¹⁵⁵ found that covalent conjugation of a B cell specific antibody to the GO nanosheets, in addition to PEGylation and adsorbed DOX, enhanced selective killing of B-cell lymphoma cells, indicating that selectively targeting cells through surface functionalization could increase therapeutic action. Folic acid (FA) surface functionalizations have also been explored for 2D nanomaterial drug delivery, targeting folate receptors overexpressed on the vast majority of cancer cells, including breast cancer cells. Zhang et al.¹⁶³ constructed a GO nanosheet functionalized with FA molecules and sulfonic acid. The resulting FA-NGO conjugate showed higher potency for killing human breast cancer MCF-7 cells in vitro when loaded with two anticancer drugs, CPT and DOX, than bare NGO loaded with either drug alone.¹⁶³ A similar receptor-mediated drug delivery approach was also developed using a different 2D nanomaterial, covalent organic nanosheet (CON), where the FA-CON conjugate selectively delivers the anticancer drug, 5-fluorouracil (5-FU), to MDA-MB-231 breast cancer cells and triggers cell death by apoptosis (Figure 10B).¹⁶⁴

Photothermal applications of 2D nanomaterials

In addition to 2D nanomaterials’ propensity to adsorb and deliver hydrophobic drug molecules, unique electronic properties enable 2D nanomaterial drug delivery to be optically controlled, yielding a quantitative release of adsorbed drug molecules on 2D nanomaterials. To this aim, photothermal mechanisms for drug release from nanosheets (Figure 10C) has been demonstrated.^165–167

We note that photothermal heating of nanosheets can cause tissue ablation,^168,169 which is another optically-controlled therapeutic modality. Critically, GO nanosheets of certain sizes (10–100s nm) have substantial absorption coefficients in the near infrared (NIR) window.^165,168 In this optical window, biological tissues have low absorption and scattering profiles, allowing for NIR light to penetrate more deeply into tissues and cells.¹⁷⁰ This poises a therapeutic opportunity for GO nanosheets that are internalized into certain cells such as cancer cells to be irradiated with NIR light; the GO nanosheets will absorb the light, heat up, ablate, and kill the cancer cells. Hence, all that is needed for this technique is surface functionalization of the GO nanosheet for the target cells or tissues, and tuning the size of the GO nanosheet to ensure internalization and robust heating at the irradiated wavelength.^168,171

In contrast, for photothermal drug release mechanisms, light irradiation does not induce tissue ablation but rather heats the GO nanosheets to induce drug unbinding due to enhanced thermal fluctuations^166,167,172 or disrupt endosomes to release encapsulated GO nanosheets.¹⁶⁵ Other 2D nanomaterials besides GO nanosheets also have strong absorbance in the optical window and have further been studied for photothermal heating. For instance, a PEGylated MoS₂ nanosheet (MoS₂-PEG) loaded with DOX was investigated for a combined photothermal- and chemotherapy,¹⁷³ where the MoS₂-PEG/DOX system in combination with NIR irradiation exhibited high antitumor effect and low cytotoxicity both in vitro and in vivo, with statistically significant reduced tumor volume in comparison to the MoS₂-PEG/DOX system without NIR irradiation. Another example of NIR-mediated photothermal heating and drug release of MoS₂ is the work of Deng et al.¹⁷⁴ who used BSA-exfoliated MoS₂ nanosheets loaded with an anticancer drug resveratrol to target human Burkitt’s lymphoma cell line Raji cells.

2D nanomaterials with liposomes for drug delivery

Besides surface passivation and functionalization for drug delivery, 2D nanomaterials can be encapsulated or adsorbed onto lipid micelles and liposomes for increased biocompatibility. Lipid micelles and liposomes are created by amphiphilic phospholipids self assembling in solution to either form monolayers with hydrophobic cores, micelles, or bilayer spheroids with hydrated cores, liposomes.¹⁷⁵ Liposomes are often used in combination with 2D nanomaterial photothermal application, heating the nanomaterial to cause liposome rupture and release of encapsulated and sequestered drugs at specific locations^176,177 (Figure 10D). In addition, nanomaterial-coated liposomes are internalized by cells through endocytosis more effectively than bare liposomes,^178,179 allowing for the passive physiological release of their encapsulated drug cargos. Lastly, layer-by-layer (lbl) nanoparticle constructs are typically composed of alternating 2D nanomaterials and lipid micelles or liposomes containing hydrophobic drugs.^176,180,181 These lbl particles enable the targeted delivery and release of drugs through several modalities including photothermal^176,180 and pH-responsive¹⁸¹ mechanisms.

Cell internalization and clearance of 2D nanomaterials

With adequate surface modification, nanomaterial biocompatibility greatly increases, and 2D nanomaterials no longer lyse the cell membranes but are internalized by cellular endocytosis, dependent on nanosheet size and cell type (Figure 11A). Linares et al.¹⁸³ found that 100 nm PEGylated GO nanosheets with fluorescein isothiocyanate (FITC) conjugation generally entered mammalian cells through macropinocytosis, the non-specific soluble uptake of extracellular fluid and solutes. However, Linares et al. additionally found cell-type specific cellular uptake with murine macrophages internalizing by clathrin-mediated endocytosis (CME), human hepatocytes internalizing by CME and phagocytosis, and Saos-2 osteoblasts internalizing by microtubule-dependent mechanisms. Nanosheet size also plays a critical role in determining the method of cellular entry. Mu et al. ¹⁸⁴ found that FITC-BSA-coated GO nanosheets exhibited size-dependent internalization in mouse mesenchymal progenitor cells, with larger nanosheets an average length of 860 nm being internalized through phagocytosis, and smaller nanosheets an average length of 420 nm being internalized through clathrin-mediated endocytosis. Lastly, we note that 2D nanomaterials do not always have to be internalized by the cell in order to deliver their therapeutic targets. Chen et al. ¹⁸⁵ found evidence that GO nanosheets could intercalate between membrane leaflets and used a theoretical model to predict GO-loaded drug diffusion through the lipid bilayer. Chen et al. ¹⁸⁵ further observed synergistic anticancer responses when the tyrosine kinase inhibitor vandetanib was absorbed onto GO nanosheets and administered to a cancer cell line. However, it was unclear if this therapeutic response was a result of GO nanosheet internalization through endocytosis mechanisms or passive drug diffusion while the nanosheet was intercalated in the membrane. Further MD simulations are warranted and expected to help unveil its mechanism in the near future.

Figure 11:

Mechanisms of cell internalization of 2D nanomaterials and carrier clearance by macrophages. (A) 2D nanomaterial cell internalization mechanisms. Reprinted from ref.,¹⁵⁰ Copyright 2021, with permission from Elsevier. (B) (Left) Nanocarrier clearance mechanism through SR-A1 recognition of denatured albumins on polymersomes (PSs). (Right) Macrophage uptake of PS-BSA after pre-treatment of an SR-A1 competitor ligand, fucoidan. Macrophage uptake of OH PS-BSA and MeO PS-BSA decreases with fucoidan, suggesting that macrophages use SR-A1 to recognize PS with denatured BSA. Adapted from ref.¹⁸⁶

The performance of a nanocarrier for drug delivery relies not only on its drug loading capacity, target specificity, and drug release mechanism but also on its circulation time. It was found that the protein corona on the nanocarrier can influence its recognition by the immune system. The study used a self-assembled soft polymeric vesicle, polymersomes (PSs), as a model system and investigated different surface functional moieties affecting the conformation of adsorbed BSA and the clearance of PSs by macrophages. The results show that BSA conformation remains intact on the phosphate-functionalized PS (Phos PS) but partially unfolds on the methoxyl- and hydroxyl-functionalized PS (MeO PS and OH PS, respectively). As the structurally altered forms of albumin are non-canonical ligands of the macrophage class A1 scavenger receptor (SR-A1), changes in the BSA conformations on MeO PS and OH PS results in their clearance by macrophage endocytosis while the intact BSA on the surface of Phos PS evades SR-A1 recognition (Figure 11B).¹⁸⁶ Although this study was performed on PS, 2D nanomaterials also have strong protein adsorption ability⁶⁶ and BSA is used as an exfoliating and stabilizing agent for GO and MoS₂ nanosheets⁶⁴ as well as a common approximation to the protein corona. Hence, it will be interesting to study how chemical modifications and conformational changes of the adsorbed protein corona affects recognition by macrophages.

Discussions and Outlook

Complementary to in vivo and in vitro experiments, in silico studies have revealed complex interactions between various 2D nanomaterials and biomolecules. Hydrophobic nanomaterials such as graphene, GO, MoS₂, and h-BN interact strongly with hydrophobic protein residues, phospholipid membranes, and nucleic acids. Protein adsorption onto 2D nanomaterials can cause structural changes and localized unfolding, particularly for α-helices. 2D nanomaterials can penetrate cell membranes and extract phospholipids, leading to membrane disruption. Nucleic acids form stabilizing π-π interactions differentially for different 2D nanomaterials; this enables the design of complex nanomaterial heterostructures for nucleic acid handling and sequencing applications. Although many bare nanomaterial interactions with biomolecules may be disruptive, 2D nanomaterial-biomolecule interactions can be modulated through surface functionalization for the safe application in nanomedicine. Given broad, hydrophobic surfaces, 2D nanomaterials can readily adsorb hydrophobic therapeutics and be used as drug delivery platforms. The unique propensity of GO nanosheets to release therapeutics in acidic environments such as tumors, holds great potential for chemical-mediated targeted therapeutic release. 2D nanomaterial optical properties, especially the inherent ability to absorb light in the near infrared window, enable an additional method of targeted delivery through photothermal drug release, as well as enable a novel targeted therapeutic modality through photothermal ablation. However, optimal application of 2D nanomaterials for drug delivery requires extensive and careful investigation of 2D nanomaterial-biomolecule interactions in vivo, which could be expensive or out of reach of current experimental methods. Thanks to advances in high performance supercomputers, state-of-the-art MD simulations complement experiments and allow studying of 2D nanomaterial-biomolecule interactions at the molecular level.

For 2D nanomaterial drug delivery, several open questions remain. Particularly, the exact structure and effects of combined drug-adsorbed, functionalized nanosheets with adsorbed protein corona has not been characterized. It is known that PEG molecules interact with drug molecules in intricate ways, such that drug adsorption is dependent upon PEG density and chain length.¹⁵⁹ However, the flexibility of the PEG chains likely changes drug loading and unloading dynamics, that in combination with interacting protein corona leads to a crowded, target-rich environment that therapeutics must escape from to selectively bind physiological targets. Surface inhomogeneity of 2D nanomaterials, including edges and defects will likely be locations of incipient binding and unbinding, arguing for the further study of wholly realistic nanomaterial systems for adequate understanding of 2D nanomaterials for drug delivery. The complexity of these systems increases even further when the nanosheets are encapsulated in lipids such as liposomes or micelles. Beyond acting as nanocarriers for small molecule drugs, 2D nanomaterial delivery of biologic therapeutics remains under-explored. Specifically, can nanomaterials adsorb biologic therapeutics, such as ssRNA in mRNA vaccines, and deliver these payloads across the cell membrane? Although being highly speculative, it is worth investigating this potential in silico as a first step. If feasible, 2D nanomaterials might become attractive candidates for delivering various biologics. Lastly, once 2D nanomaterials have delivered their therapeutics, nanomaterial degradation and excretion must be fully understood to prevent toxic accumulations in body tissues. The potential cytotoxicity of 2D nanomaterials to host cells over time also warrants further studies regarding dosage and period of exposure. Although exact physiological effects of 2D nanomaterials should always be tested by experiment, structure and dynamic modeling of 2D nanomaterials is important to help elucidate interaction mechanisms and generate hypotheses to increase comprehension of 2D nanomaterial-biomolecule interactions. The in silico studies of 2D nanomaterial-biomolecule interactions elaborated here are expected to accelerate the discovery of safe and effective nanomedicine.